The rest have ἐν βρώσει ἢ ἐν πόσει. This may be explained as a very obvious, though not very intelligent, alteration of scribes to conform to the disjunctive particles in the context, ἢ ἐν μέρει ἑορτῆς ἢ νεομηνίας ἢ σαββάτων.
In this same context it is probable that B retains the right form νεομηνίας (supported here by F G and others) as against the Attic νουμηνίας. In the same way in iii. 25 κομίσεται and iv. 9 γνωρίσουσιν B (with some others) has resisted the tendency to Attic forms.
ii. 18 ἃ ἕορακεν.
ii. 18, the omission of the negative.
That this is the oldest reading which the existing texts exhibit, will appear from the following comparison of authorities.
(1) ἃ ἑώρακεν ἑώρακεν) A B א* D*, 17*, 28, 67**; the Old Latin authorities d, e, m; the Memphitic, Æthiopic, and Arabic (Leipz.) Versions; Tertull. c. Marc. v. 19 (‘ex visionibus angelicis’; and apparently Marcion himself also); Origen (c. Cels. v. 8, I. p. 583, though the negative is here inserted by De la Rue, and in Cant. ii, III. p. 63, in his quæ videt); Lucifer (De non conv. c. hær. p. 782 Migne); the Ambrosian Hilary (ad loc. explaining it ‘Inflantur motum pervidentes stellarum, quas angelos vocat’). So too the unknown author of Quæst. ex N. T. ii. 62 in August. Op. III. Appx. p. 156. Jerome (Epist. cxxi ad Alg. § 10, I. p. 880) mentions both readings (with and without the negative) as found in the Greek text: and Augustine (Epist. 149, II. p. 514), while giving the preference to quæ non vidit>, says that some MSS have quæ vidit>.
(2) ἃ μὴ ἑώρακεν (ἑόρακεν) אc C Dbc K L P, and the great majority of cursives;
(3) ἃ οὐκ ἑώρακεν F G.
The negative is also read in g; in the Vulgate, the Gothic, both the Syriac, and the Armenian Versions; in the translator of Origen In Rom. ix. § 42 (IV. p. 665), in Ambrose In Psalm. cxviii Exp. xx (I. p. 1222), and in the commentators Pelagius, Chrysostom, Theodore of Mopsuestia (Spic. Solesm. I. p. 132 ‘quæ nec sciunt’), Theodoret, and others.
From a review of these authorities we infer that the insertion of the negative was a later correction, and that ἃ ἕωρακεν (or ἕορακεν) represents the prior reading. In my note I have expressed my suspicion that ἃ ἕωρακεν (or ἕορακεν) is itself corrupt, and that the original reading is lost.